garden
Skill1 starsChanged 2 months agoMITView on GitHub
Tend a codebase between features: find drift in one area, group it by cause, and fix the causes worth their cost in one behavior-preserving pull request. Use when asked to garden, clean up, improve architecture, find what slows the product, or stop a defect that keeps coming back.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- harness-mini/harness-minitheir own version, first seen Nov 1, 2026
- avlihachev/claude-skill-gardentheir own version, first seen Mar 28, 2026
- maragudk/skillstheir own version, first seen Feb 12, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- maragudk/skills★ 49same name
- avlihachev/claude-skill-garden★ 7same name
- harness-mini/harness-mini★ 7same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository34 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adrSkill
- agents-mdSkill
- architectSkill
- bug-huntSkill
- coding-standardsSkill
- deliverSkill
- diagnosing-bugsSkill
- howSkill
- program-designSkill
- prototypeSkill
- pruneSkill
- refineSkill
- researchSkill
- retroSkill
- setup-verificationSkill
- shapeSkill
- shipSkill
- skill-guideSkill
- spikeSkill
- strategic-programmingSkill
- walkthroughSkill
- whySkill
- adaptSkill
- editorSkill
- issueSkill
- pr-descriptionSkill
- pr-reviewSkill
- writing-for-peopleSkill
- grillingSkill
- unslopSkill
- wait-whatSkill
- writing-for-agentsSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 37
Every picked cause is fixed at its root in a pull request with green checks, each new check goes red on its violation, behavior is …
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Garden
Agents copy whatever the codebase already does, good or bad, so drift compounds. Gardening runs alongside feature work to keep the codebase a place where the right change is the easy one.
Scan first and fix later: a queue of findings shows the causes that fixing them one by one would hide. Run to the end on your own; anything for the human goes in the report, written in the user's active language. Write it and every issue you file with writing-for-people.
Lenses
Pick the lens the user names, or the one with the strongest signal:
- Architecture: concepts owned in several places, shallow modules that only pass calls through, files that always change together, boundaries the code crosses against its design. Use the deletion test from
strategic-programming. - Consistency: deviations from the project's agent instructions and coding standards, the same mechanism implemented several ways, the same defect fixed in more than one feature, dead code and leftovers.
- Performance: product budgets and the work behind them, such as round trips, queries, bundle size, and renders, measured before and after.
- Tests: run
prune. - Verification: run
setup-verification.
Process
- Scope. Take the area the user names, or the one with the strongest signal: churn hot spots in history, defects that came back, follow-ups listed in recent pull requests, slow paths. Prefer an area no recent gardening covered: every pull request and issue gardening opens carries the
gardenlabel, created when the project lacks it. State the choice. - Scan into a queue. Record each finding with its evidence; change nothing yet.
- Group by cause. Many findings usually share one. For each cause choose the strongest structural fix in the order
strategic-programminggives. - Pick. Rank the causes by leverage and fix the ones worth their cost, unless the user picks. A cause whose fix would change behavior, a contract, or a decision an ADR records is not gardening: it goes to step 6.
- Fix the picked causes on one branch off the base, in one pull request, each cause in its own commits. Behavior stays the same: when the affected code lacks tests that would catch a change, add characterization tests first. When the run also prunes, because the user asks or the tests lens leads, run
prunein a fresh agent on its own worktree off the same base and bring its commits onto the branch. Then runrefineandship. - File the rest. Search open issues first, then file each unpicked cause as one issue, written with
issue, with its instances and what it costs. A finding that looks like a defect, such as a swallowed error or a contract the code breaks, is filed as a suspected bug forbug-huntto confirm.
Report
End with the area and why, the pull request and its check state, each cause fixed, each issue filed, and what was not run.
Done
Every picked cause is fixed at its root in a pull request with green checks, each new check goes red on its violation, behavior is unchanged, every other cause is filed or already has an issue, and the report is delivered.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

